AN ABSTRACT OF THE THESIS OF

Mark Tower Smith for the degree of Master of Science

in Mechanical Engineering presented on November 26, 1984

Title: EXPERIMENTAL DEVELOPMENT AND CHARACTERIZATION OF A SMALL-SCALE

PLANAR FLOW CASTING (PFC) MACHINE

Abstract Approved:

Redacted for Privacy

Murli Saletore

The design, fabrication, and operation of a relatively simple

planar flow casting (PFC) machine optimized for small-batch processing

is described. Several design features found beneficial to PFC process

operation include a ground nozzle stopper to retain the alloy charge

during melting; a remote, large-volume pressure vessel connected to

the crucible gas system to reduce temperature induced pressure

fluctuations; and the use of a nested induction coil that allows both

the melt charge and the crucible reservoir to be located close to the

cooling wheel.

A partial factorial experimental approach was used to evaluate the

single-variable and variable-interaction effects on ribbon thickness

for 28 PFC process runs in which the cooling wheel surface velocity

(Vs), crucible ejection pressure (P

e), and crucible nozzle clearance

gap (Gs) were varied.

Linear regression analysis techniques were used to determine

single-variable equations relating ribbon thickness to each of the

independent process variables. In addition, multiple linear regres-

sion analysis was used to develop a multi-variable equation for

ribbon thickness in which both single-variable and variable interac-

tion effects are included.

Examination of the rapidly solidified, Fe-based ribbons for dimen-

sional uniformity, atomic structure, and cooling wheel wetting pattern

indicates that good quality glassy ribbon can be produced with proper

selection of the independent process variables.

In particular, it was found that relatively high values of cooling

wheel surface velocity and crucible ejection pressure result in improved

wetting of the cooling wheel surface and promote a more rapidly solidi-

fied, glassy, ribbon structure. Variations in the crucible nozzle-

clearance gap were found to have little effect on cooling wheel surface

wetting or ribbon atomic structure within the experimental range

utilized.

EXPERIMENTAL DEVELOPMENT AND CHARACTERIZATION OF ASMALL-SCALE PLANAR FLOW CASTING (PFC) MACHINE

by

Mark T. Smith

A THESIS

submitted to

Oregon State University

in partial fulfillment ofthe requirements for the

degree of

Master of Science

Completed November 26, 1984

Commencement June 1985

APPROVED:

Redacted for Privacy

Assistant Professor of Mechanical Engineering in charge of major

Redacted for Privacy

Head of Departmen t Mechahltal Engineering]

'Redacted for Privacy

Dean of Graduate

rn

fol

Date thesis is presented November 26, 1984

Typed by Mayo Poquette for Mark T. Smith

ACKNOWLEDGMENT

As in any area of research, contributions to the final published

product come from a diverse group of people working in a variety of

disciplines and occupations. Such is the group of people I wish to

acknowledge for their many important contributions.

The first person I want to thank is Murli Saletore, my advisor, for

the many contributions and discussions which he provided. His interest

in this work and the comfortable working relationship that developed

were greatly appreciated.

This work was supported by the Materials Section, Bureau of Mines,

Albany Research Center, under the direction of H. W. Leavenworth.

Altheuah there are too many contributing individuals at the Bureau to

list in entirety, a number require special mention. I am particularly

grateful to L. G. McDonald and G. A. Fortier for their direct

involvement and many contributions. I thank G. Asai, S. R. Brooks,

J. M. Burrus, and R. C. Doan of the Materials Section for their many

technical contributions. I am indebted to R. H. Maier for his skilled

and innovative glass work, and likewise to R. A. McCune for the x-ray

diffraction analysis which was so important to this work. Finally, I

thank M. L. Farlee, S. A. O'Hare, and P. A. Romans for a variety of

helpful contributions.

A special thank you goes to J. C. Rawers for his help and guidance

in the design of the statistical experimentation program utilized for

this work, as well as the many helpful discussions we had.

Finally, I thank M. W. Poquette for contributing her typing and

editing skills, which did much to improve the presentation of this work.

TABLE OF CONTENTS

1.0 INTRODUCTION/LITERATURE REVIEW 1

2.0 PROCESS DEVELOPMENT 6

2.1 Crucible Design and Fabrication 6

2.2 Crucible Gas System 11

2.3 Induction Heating System 12

3.0 EXPERIMENTAL PROCEDURE 15

3.1 Master Alloy Preparation 15

3.2 Statistical Characterization Program 16

4.0 RESULTS 25

4.1 Regression Analysis of Single Variable EffectsOn Ribbon Thickness 25

4.2 Results of Multiple Linear Regression Analysis . . 32

4.3 Results of X-Ray Diffraction Analysisof Ribbon Structures 33

4.4 Results of Qualitative Evaluation of RibbonDimensional Quality 35

4.5 Results of Ribbon Cooling Wheel Surface ContactPattern Studies 36

5.0 DISCUSSION OF RESULTS 37

5.1 Effects of Process Variables on Ribbon Thickness . 37

5.2 Effect of Process Variables on Ribbon CoolingSurface Pattern 39

5.2.1 Effect of Cooling Wheel Surface Velocity(Vs) 40

5.2.2 Effect of Ejection Pressure (Pe) 44

5.2.3 Effect of Nozzle Clearance Gap (Gs) 46

6.0 CONCLUSIONS 51

BIBLIOGRAPHY 53

APPENDICES 54

FIGURES

1. Diagram Showing Physical Layout of the Melt Crucibleand Cooling Wheel 3

2. Plan View of Nozzle Slot Opening Showing Nozzle SlotLength (L) and Nozzle Slot.Width (Wn) Dimensions 3

3. Schematic Diagram of the PFC Process Machine 7

4. Planar Flow Casting Machine Shown During Operation 8

5. Photograph of a Typical PFC Crucible Showing MajorDesign Features 10

6 Photograph of Experimental Induction Coils Used forMelting. The nested coil configuration is on the left,with the conventional straight wound coil shown on theright for comparison 14

7. Photograph of Prealloyed Melt Specimens (On Left)Shown Beside Arc-Melted Button. Material Compositionis Fe

78B13

Si9% a/o 17

8. Photomicrograph of Arc-Melted Fe78 B13 Sig a/o Button(400X) 18

9. Schematic Diagram of Single Variable ExperimentalTest Points 20

10. Schematic Diagram of Variable Interaction ExperimentalTest Points 20

11. Ribbon Specimens Used to Determine Qualitative Ratingsfor Dimensional Uniformity 24

12. Linear Curve Fit for Ribbon Thickness (S) as a Functionof Crucible Ejection Pressure (Pe) ?9

13. Linear Regression Power Fit Curve for Ribbon Thickness (S)as a Function of Cooling Roll Speed (Vs) 30

14. Exponential Curve Fit for Ribbon Thickness (S) as aFunction of Crucible Nozzle Clearance Gap (Gs) 31

15. Examples of X-Ray Diffraction Patterns Obtained forEach of the Structural Groups (a) Amorphous, (b) PrimarilyAmorphous, and (c) Partially Crystalline 34

16. Cooling Wheel Surface Contact Pattern for CommercialMetallic Glass (MG 2605), 50X 41

FIGURES (cont'd.)

17. SEM Photograph of Cooling Wheel Surface Contact Patternfor Commercial Metallic Glass (MG 2605), 120X 41

18. Ribbon Cooling Wheel Contact Surface Produced at High Vs

(23.6 m/s), 50X 42

19. Ribbon Cooling Wheel Contact Surface Produced at Low Vs

(9.97 m/s), 50X 42

20. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at High Vs (23.6 m/s), 120X 43

21. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at Low V

s(9.97 m/s), 120X 43

22. Ribbon Cooling Wheel Contact Surface Produced at High Pe

(17.2 KPa), 50X 45

23. Ribbon Cooling Wheel Contact Surface Produced at Low Pe

(6.9 KPa), 50X 45

24. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at High Pe (17.2 KPa), 120X 47

25. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at Low P

e(6.9 KPa), 120X 47

26. Ribbon Cooling Wheel Contact Surface Produced at High Gs

(0.381 mm), 50X 48

27. Ribbon Cooling Wheel Contact Surface Produced at LOW G(0.152 mm), 50X

s48

28. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at High Gs (0.381 mm), 120X 50

29. SEM Photograph of Ribbon Cooling Wheel Contact SurfaceProduced at Low G

s(0.152 mm), 120X 50

TABLES

I Results of Chemical Analysis of a Randomly Chosen RibbonSample. Nominal Composition is Provided for Comparison. . . 19

II Results of Single Variable Experimental Runs 26

III Results of Interaction Effects Experimental Runs 27

IV Average Percent Variation and Standard Deviation of ThicknessOver the Length of Ribbon Specimens 28

V X-Ray Diffraction Results for Cooling Wheel and Free SurfacesOf Selected Ribbon Specimens 35

EXPERIMENTAL DEVELOPMENT AND CHARACTERIZATION OF A SMALL-SCALE

PLANAR FLOW CASTING (PFC) MACHINE

1.0 INTRODUCTION/LITERATURE REVIEW

The Planar Flow Casting (PFC) process(1)

is one of the more

recent liquid-quenching rapid solidification techniques to be

developed for casting of microcrystalline and amorphous strip

materials.

First reported by M. C. Narasimhan in the PFC patent document,(1)

the process produces continuous metal strips by forcing liquid metal

through a nozzle with a rectangular cross-section onto the surface of

a rotating cooling roll. The ribbon, which solidifies at the cooling

wheel-melt puddle interface, is carried off by the rotating cooling

wheel in the form of a long, thin strip.

The cooling rate, T, to which the ribbon or strip is subjected,

has been shown to be controlled largely by ribbon thickness.(2) As a

result, knowledge of the effects of process variables and equipment

configuration on strip dimensions is essential for optimum operation

of the PFC process.

Under stable ribbon casting conditions, with no melt splatter or

ribbon discontinuities, an overall mass balance exists such that the

melt flow rate from the nozzle equals the rate of ribbon formation.

A number of researchers(3

'

4'

5'

)

have proposed empirical relationships

among ribbon thickness (S), melt flow rate (Q), and the cooling wheel

surface velocity (Vs) for both the Chill Block Melt Spinning (CBMS)

and PFC processes. For the PFC process it has been shown that S is

proportional to Q/Vs. This approach assumes a knowledge of how the

independent process control variables (such as crucible ejection

pressure (Pe), crucible nozzle clearance gap (Gs), and nozzle slot

width (Wn) affect Q.

2

Other researchers have employed photographic techniques to

examine melt puddle geometry during CBMS and PFC processing.(4

'

6'

7)

This has resulted in empirical relationships between S, the melt

puddle length (1), and the cooling substrate velocity (Vs). In

general, these relationships are of the form Scc em, where 0 is the

"contact time" defined as 1/Vs

. This approach has been of some use in

the study of ribbon formation models, but provides only indirect

information about the controllable process variables which affect melt

flow.

Relatively few studies have been reported in which the effects of

the primary controllable process variables on ribbon thickness are

considered. Narasimhan(1)

identified four process variables which

were found to have a significant effect on ribbon thickness. Included

are the previously mentioned variables Vs , Pe , and Gs

plus the

crucible nozzle geometry which consists of the nozzle slot width (Wn)

and the widths of the leading and trailing nozzle lips. A diagram

showing the physical location of the PFC crucible is shown in

Figure 1. General features of the crucible nozzle slot are shown in

Figure 2.

Although Narasimhan does not present empirical or theoretical

relationships between the variables, he does provide a qualitative

description of the effects of the individual variables on ribbon

thickness. In particular, Wn was found to be an important factor in

controlling melt flow rate only between certain limiting values. For

nozzle slot widths between 0.5 mm and 1.0 mm it was found that the

nozzle clearance gap was the main parameter affecting melt flow rate.

In this situation, casting was found to be gap controlled, while for

Wn<0.5 mm, the melt flow rate was increasingly limited by the nozzle

slot opening dimension. Huang,(8)

in an experimental study on the

effects of melt delivery conditions on ribbon formation, proposed a

similar relationship between Wn and Gs. In his work, a relatively

large crucible nozzle clearance aap was used with the result that

casting was controlled primarily by nozzle slot dimensions. Huang

3

SecondaryMelt Reservoir

Nozzle ClearanceGap Gs

6.)

Cooling Wheel

CrucibleNozzle

Crucible MeltReservoir

Figure 1. Diagram Showing Physical Layout of the Melt Crucible and Cooling Wheel

Directionof

Wheel Rotation

Cc-

L

Figure 2. Plan View of Nozzle Slot Opening Showing Nozzle Slot Length (L)and Nozzle Slot Width (Wr,) Dimensions.

4

noted that for slot controlled casting, ribbon thickness is propor-

tional to the square root of the ejection pressure, Pe. However, as

Wnbecomes increasingly larger than G

s, casting approaches gap

controlled conditions described by Narasimhan and the relationship is

no longer applicable.

In a related study by Huang and Fiedler,(4)

the effects of

cooling wheel velocity, Vs, on the ribbon-wheel interfacial thermal

conductance (h) and ribbon contact area are reported. It was found

that h, calculated from a continuous casting solidification model,

increases with increasing values of cooling surface velocity. The

improvement in h at higher Vs values was attributed to a similar

increase in ribbon contact area at the higher substrate velocities.

Since conductive heat transfer is the dominant mechanism of

cooling in chill-block casting,(2)

optimization of the ribbon contact

area is an important step in the process of attaining higher cooling

rates. When processing is carried out in ambient atmospheric

conditions, a portion of the boundary layer air (adjacent to the

rotating cooling wheel) becomes entrapped between the melt puddle and

the cooling wheel, leading to the formation of air pockets and

consequently of regions with poor heat transfer rates. Huang (9) found

that both the air pocket distribution and morphology were affected by

the surface finish of the cooling wheel. In his study, ribbon was

cast on cooling wheels having either a smooth, polished surface or a

uniformly rough, "matte" finish. Examination of the ribbon by

scanning electron microscopy (SEM) and by surface profilometry found

that the "matte" wheel surface resulted in a more uniform distribution

of air pockets and a higher estimated value of the heat-transfer

coefficient (h).

It should be noted that the PFC process work reported by Huang

and Fiedler was conducted on relatively large-scale processing

equipment utilizing melt charges of several Kg. Such melt quantitites

generally require water cooling of the quench roll, a brush or wiper

5

to maintain cooling wheel surface condition, and often some form of

active control of the crucible clearance gap.(10)

This type of

equipment is typically used to produce several hundred meters of strip

or ribbon at a time, under relatively stable casting conditions. For

many workers in the Rapid Solidification Technology (RST) field who

need comparatively small quantities of ribbon for research purposes,

such equipment is needlessly complex and expensive.

The purpose of the present work is to report on the development

and characterization of a simple, small-scale, PFC processing machine.

The design and development of the process equipment will be described

with emphasis on those features determined to be relevant for

small-scale processing. A process characterization program exploring

the effects of the primary process variables on ribbon thickness and

contact area is also reported.

6

2.0 PROCESS DEVELOPMENT

The Planar Flow casting process was chosen for the present work

primarily because of the ability to produce amorphous ribbon in widths

which exceed those obtainable with other chill-block casting

processes. (3) Although the PFC technique has become a commercially

prominent process, its use in RST research has been limited. This

appears to be due in part to the limited process information available

in the literature. Whereas a number of papers have been presented on

the construction and operation of CBMS process equipment,(11,12,13)

the PFC patent document has, until recently, been the primary source

of information for PFC equipment design and configuration.

The following is a general description of the PFC equipment used

for this work. In addition, the design of the crucible and melting

system will be described in some detail, with emphasis on those

features found beneficial to small batch-quantity PFC processing.

A schematic diagram of the PFC process machine is shown in

Figure 3. Major components include a 30.5 cm dia copper cooling wheel

(A), variable speed cooling wheel drive system (B), quartz furnace

crucible assembly (C), the adjustable crucible mounting frame (D),

argon gas pressure system (E) with equalizing vessel (F), induction

coil (G), and the machine frame (H). Instrumentation consists of a

200-2000 rpm direct drive tachometer (I) mounted to the drive shaft

end, a 0-10 psig pressure gage (J) to monitor the crucible ejection

pressure, and a Leeds and Northrup single color pyrometer (K) for

measurement of melt temperature. A photograph of the PFC process

machine during operation is shown in Figure 4.

2.1 CRUCIBLE DESIGN AND FABRICATION

The primary functions of the crucible system are to provide a

reservoir in which the alloy charge is melted and, once molten,

deliver the melt through the crucible nozzle onto the rotating

cooling wheel. This requires that the crucible be constructed of a

(C) CrucibleAssembly

(G) InductionCoil

To InductionPower Unit 11*--4

(H) MachineFrame

(A) CopperCooling Whl

(B) CoolingWhl. DriveSystem

(D) CrucibleMounting Frame

Front View

Figure 3. Schematic Diagram of the PFC Process Machine.

(E) Argon GasSystem

(K) OpticalPyrometer

(J) Gas PressureGage

To RegulatedArgon Source

(I) Tachometer

(F) PressureEqualizing Vessel

8

Figure 4. Planar Flow Casting Machine Shown During Operation.

9

material capable of withstanding the rapid temperature changes

associated with the induction melting process,(12)

and preferably have

little reactivity with the alloy in the molten condition.

All crucibles used for this work were constructed of fused silica

(quartz) and feature an integral rectangular slotted crucible nozzle.

Fused silica, in addition to providing the necessary thermal

properties required for melting Fe-based alloys, has good hot forming

characteristics and provides a transparent vessel suitable for

observation of the melting process.

Fabrication of the crucible nozzle to the dimensions recommended

by Narasimhan (1) was accomplished by hot-forming a quartz tube around

a tapered, rectangular, graphite mandrel. Although this technique

resulted in some variation in nozzle dimensions, careful grinding of

the nozzle end allowed the slot width to be held within the desired

dimensional range (0.5 mm< Wn<1.0 mm) .

To optimize the crucible design for small-batch PFC processing, a

ground quartz ball and socket was incorporated into the design in

order to provide positive retention of the melt under pressure. This

feature, which does not appear to have been previously reported in the

literature, allows the alloy charge to be melted and stabilized at a

desired liquidus temperature while under a constant preset ejection

pressure. In addition, the quartz socket provides a reduced diameter

tube from which a small secondary melt reservoir is fabricated. The

secondary reservoir, located below the ball and socket assembly,

serves to collect the melt directly above the tapered nozzle slot

opening and provide continuous delivery of the melt to the cooling

wheel.

A typical PFC crucible used in the experimental work is shown in

Figure 5. Major features are the 31 mm diameter crucible tube (A) to

which the ground quartz socket (B) with integral secondary reservoir

(C) and nozzle (0) are fused. Shown beside the crucible is the ground

10

(A) Crucible Tube

(B) Ground Quartz

(E) Ground Ball StopperSecondaryReservoir

Figure 5. Photograph of a Typical PFC Crucible Showing Major Design Features

11

quartz ball stopper (E) which is joined to the stopper rod (F). The

ground ball stopper assembly seats into the quartz socket (B)

providing a positive seal between the crucible and nozzle during

melting. The stopper rod (F) passes through the o-ringed crucible

bell top (G) allowing external control of the stopper assembly

position.

Application of a thin film of high temperature silicone grease to

the stopper ball, stopper rod, and bell top flange provides a

gas-tight seal during melting. To prevent cracking of the crucible

during heating of the pre-alloyed melt charge, the melt specimen (H)

is suspended freely within the crucible by means of a tungsten wire

(I). As the melt charge approaches liquidus temperature it drops from

the higher melting tungsten wire and forms a molten puddle around the

crucible stopper.

2.2 CRUCIBLE GAS SYSTEM

The argon gas system used to provide the crucible ejection

pressure should, ideally, supply a constant pressure throughout the

casting run. The system that has evolved from this study consists of

a regulated argon source connected directly to the furnace crucible

with a separate in-line pressure gage for monitoring the static

(stopper ball in the sealed position) crucible pressure. In addition,

a remote 15.0 z sealed pressure vessel is connected in parallel with

the crucible gas system. This last feature was found to be effective

at minimizing pressure fluctuations caused by heating of the argon gas

during melting operations and provided a relatively constant ejection

pressure during the casting run.

Development runs performed prior to fitting the remote pressure

vessel showed substantial variations in ribbon thickness between runs

using the same statically preset ejection pressure. Investigation

revealed that the ejection pressure was highly dependent on the time

required to melt the alloy charge. Longer melting times resulted in

an increase in argon gas temperature and, therefore, an increased

12

static ejection pressure. The addition of a comparatively large

volume remote, pressure chamber (VR) connected to the crucible gas

chamber volume (VC) reduces the initial pressure fluctuation (APint

)

by the ratio of the two volumes.

AP = APint

VC

VR + VC

(1)

Where AP = observed pressure change with remote reservoir

Where APint

= initial pressure change without remote reservoir

2.3 INDUCTION HEATING SYSTEM

One of the major differences between the PFC and CBMS process

involves the location of the crucible. Whereas melt spinning

crucibles are mounted well above the cooling wheel, PFC crucibles are

located close to the cooling wheel, as shown in Figure 1, requiring

careful design and sizing of the induction coil heating system. In

addition, it was found that efficient induction melting of the

relatively small melt charges utilized for this work required precise

location of the melt specimen in the induction field.

The induction unit used for this work was a Scientific Electric

Highboy 12.5 kW induction heating generator, Model S.12.5. Typical of

older generation induction equipment, this unit uses air-cooled triode

vacuum tubes in a grid-plate feedback oscillating circuit. Normal

operating frequencies are 200 kHz to 600 kHz. Melting experiments

indicated that an induction coil with approximately 15 coil turns at a

mean coil diameter of 38 mm (dictated by the crucible design) provided

good coupling with the melt charge.

Constructed in the conventional manner with the coil turns on a

uniform diameter, the resulting coil length (in the axial direction)

was approximately 125 mm. Although such a coil was used during

13

initial experimental runs, it was found that coupling between the

induction field and the melt charge was inconsistent. When the melt

charge was suspended on the tungsten wire above the crucible bottom,

coupling was good and the melt specimen heated rapidly. However, as

the melt charge approached the liquidus temperature and dropped to the

bottom of the crucible, coupling decreased and the desired melt

temperature could not be obtained. Subsequent investigation indicated

that as the melt charge dropped to the crucible bottom, it moved

sufficiently away from the longitudinal center of the coil to decrease

the induction coupling. If the melt charge was re-centered along the

coil length, coupling was restored and heating continued. Because of

the physical geometry of PFC crucibles, which locates the crucible

nozzle and molten charge close to the cooling wheel, repositioning of

the induction coil to center the charge in the induction field was not

possible because of the danger of arcing between the coil and cooling

wheel.

In order to solve this problem, a second induction coil was

fabricated using a "double-deck" or nested configuration. With the

nested configuration, eight of the required fifteen coil turns were

wound axially at a mean coil diameter of 38 mm, while the remaining

seven coil turns were wound concentrically around the inner eight

coil turns at a mean diameter of 53 mm separated by wood insulating

strips. A photograph of the nested induction coil is shown in

Figure 6 with the conventional straight-wound coil for comparison.

The primary advantage of the nested coil is that the overall coil

length is reduced to approximately one-half the length of the

conventional straight-wound coil. This allows the induction coil to

be positioned with the melt charge more nearly centered along the coil

axis, while maintaining a sufficient distance between the coil and the

cooling wheel to prevent arcing problems. With the nested coil

installed, good coupling between the induction field and the melt

charge was obtained throughout the melting operation.

14

Figure 6. Photograph of Experimental Induction Coils Used for Melting. The Nested CoilConfiguration is on the Left, With the Conventional Straight Wound Coil Shown onthe Right for Comparison.

15

3.0 EXPERIMENTAL PROCEDURE

The objective of the experimental characterization program was to

determine the effects of the major controllable process variables on

both ribbon thickness and structure. To ensure the validity of the

statistically designed experimental program, emphasis was placed on

minimizing variations in the prealloyed melting specimens and non-

variable processing conditions. Also described in this section are

analytical procedures used to evaluate ribbon specimens produced

during process characterization experiments.

3.1 MASTER ALLOY PREPARATION

All master alloy materials used for experimental runs were

prepared by vacuum arc melting stochiometric mixtures of laboratory-

purity elements. The Fe-based alloy selected for the statistical

experimentation program had the following composition: Fe7RB13Si9 a/o.

This alloy is reported to form a ductile amorphous structure over a

wide range of ribbon thicknesses(15)

and has the same nominal

composition as the commercial Metglas 2605-S2 alloy, allowing

comparison of x-ray diffraction results and ribbon contact patterns.

Prealloyed melting material was prepared by accurately weighing high

purity electrolytic iron (99.9%), crystalline boron (99.8%), and

silicon (98.6%) to a total button weight of 100 g. Total button

weight was recorded before and after melting to determine any weight

loss due to the melting process.

The arc melting procedure consisted of loading three 100 g

alloyed button charges into a water-cooled copper hearth plate, at

which time the arc furnace was sealed. Vacuum was then applied to the

furnace chamber followed by a series of helium gas back-fills. All

melting was carried out under a positive helium pressure to prevent

vaporization of the boron and silicon. Each 100 g button was then

melted three to five times with the button being turned over after

each melt to ensure homogeniety of the elements.

16

At the completion of the melting process, individual buttons were

weighed and those showing a weight difference of more than 0.5 g were

rejected.

To prepare individual alloy charges for experimental runs, the

buttons were sectioned transversely into 8-12 g pieces using a liquid

cooled abrasive sectioning saw. Both faces of the button pieces were

examined for inclusions, and several button ends were mounted as

metallographic specimens and examined optically for structural

uniformity.

Following abrasive sectioning, the individual button sections

were drilled at one end for placement on the tungsten crucible wire.

A typical 100 g button and several drilled button sections are shown

in Figure 7, while the microstructure of an arc-melted button is shown

in Figure 8.

In addition, a randomly chosen sample of cast Fe78813Si9 a/o

ribbon was analyzed for composition, and the results are listed in

Table I.

3.2 STATISTICAL CHARACTERIZATION PROGRAM

Following completion of the development runs, a statistically

designed experimental program was conducted to determine the effects

and optimum operating values for the three controllable process

variables, cooling wheel surface velocity (Vs), crucible ejection

pressure (Pe), and nozzle clearance gap (Gs). The dependent variable

chosen for this work was ribbon thickness (S) as measured by means of

a micrometer.

For each independent process variable, a low, medium, and high

experimental value was selected based on results of the development

runs, machine limitations, and published literature. A partial

factorial experiment design was used to form an experimental matrix

consisting of seven test points for determination of first order

17

Figure 7. Photograph of Prealloyed Melt Specimens (On Left) Shown Beside Arc-MeltedButton. Material Composition is Fe79 B13 Sig a/o.

18

200 pm St

Figure 8. Photomicrograph of arc-melted Fe-781313819 a/o button, (400X).

19

Table I

Results of Chemical Analysis of a Randomly Chosen Ribbon Sample

Nominal Composition is Provided for Comparison

Element a/o Fe

79.0

78.7

B

13.0

12.8

Si

9.0

8.5

0

0.030 0.035

Nominal

Analysis

effects and eight test points for determination of second order and

interaction effects. Graphical representation of the experimental

test space is shown in Figure 9 for the seven single variable test

points, and Figure 10 for the eight higher order test points. The

mid-value test point was repeated for each series of experimental

runs; a total of five data sets was obtained for this data point.

Four additional experimental runs were made at extreme values of Vs,

Pe

, and Gsto extend the range of the single variable data sets

(points 18, 19, 20, and 21, Figure 9). The higher order test points

(Figure 10) were run in two experimental series with the mid-value

test point and four higher order test points (11, 12,16, 17,

Figure 10) being repeated. Duplicated test points were used to

determine repeatability of the process equipment and statistical

variance of the experimental program. In total, 28 experimental runs

were completed during the characterization program.

For all runs, the cooling wheel surface condition, melt tempera-

ture, and equipment configuration were held constant. The crucible

nozzle slot width (Wn

) was measured prior to each experimental run.

When the slot width exceeded approximately 0.9 mm, the crucible was

replaced with a crucible having a minimum slot width of 0.5 mm. With

this range of nozzle slot (0.5mm<Wn<0.9mm), Narashimhar

(1)has

reported that the melt flow rate is controlled primarily by the

20

Figure 9. Schematic Diagram of Single Variable Experimental Test Points.

Figure 10. Schematic Diagram of Variable Interaction Experimental Test Points.

21

nozzle clearance gap Gs. Since Wnwas not considered an independent

variable in the experimental program, it was necessary to maintain

this range of slot widths to ensure that melt flow was controlled by

nozzle clearance gap.

The following procedure was used to prepare the PFC equipment for

all process runs:

A pre-alloyed melt specimen weighing 8 - 12 g was wire brushed and

sanded to remove any oxide, then ultrasonically cleaned in alcohol and

dried. The crucible stopper ball was then coated with a thin film of

silicone grease and the specimen suspended from the stopper by means

of a tungsten wire. After loading the alloy specimen and stopper into

the crucible, the crucible bell top was then greased and sealed to the

crucible. The crucible was then flushed several times with argon gas

and the desired crucible ejection pressure achieved by means of the

argon regulator. The ejection pressure, which was monitored by means

of the in-line pressure gage, was allowed to stabilize for approximately

ten minutes prior to the run. Prior to each PFC run, the cooling

wheel surface was refinished with 400 grit paper with the cooling

wheel rotating at 150 - 200 rpm. After sanding, the wheel was cleaned

with alcohol to remove any remaining abrasives and oil films. The

crucible nozzle gap was then adjusted using the proper size steel

feeler gage. Because of cooling wheel eccentricity (approximately

0.038 mm total indicated runout), the gap was checked several times at

different cooling wheel positions.

Melting was then started by turning the induction unit on at the

preset power level. Melt temperature was monitored by a single color

pyrometer, preset at 1275 °C target temperature. When melt brightness

reached that of the pyrometer, the crucible ejection pressure was

rechecked and the crucible stopper lifted to initiate the casting run.

Typical casting runs lasted approximately two to four seconds,

producing 9.0 m to 12.0 m of ribbon.

22

Individual runs were evaluated on the basis of ribbon thickness,

ribbon structure, ribbon quality, and cooling wheel contact pattern

in the following manner. For each experimental run, specimens were

cut from the beginning, middle, and end of the cast ribbon. Thickness

measurements were recorded and averaged for five points along the length

of the start and finish ribbon specimens, and seven points along the

middle ribbon specimens.

For each run, an x-ray diffraction specimen was prepared from the

middle ribbon section by mounting several 25 mm ribbon lengths onto a

glass specimen slide coated with double-backed adhesive tape. All x-ray

specimens were submitted with the cooling wheel ribbon surface exposed

for analysis. Initial x-ray diffraction results had shown that although

the ribbon free surface frequently was amorphous, the cooling wheel

ribbon surface often had detectable crystallization. X-ray diffraction

analysis of the cooling wheel ribbon surface, therefore, represented a

more stringent test of amorphousness.

The x-ray diffraction patterns were obtained using a Norelco

APD3600-2 instrument. This instrument consists of a constant

potential generator (Model XRG-3100) with a fine line focus Cu x-ray

tube, single crystal graphite monochrometer, scintillation detector,

and high speed electronics. Electronic control was provided by a Data

General Nova 4x mini-computer which supports the full Joint Committee

on Powder Diffraction Standards (JCPDS) data base. All patterns were

obtained at 45 kv and 40 ma from 5 degrees to 90 degrees two theta

(2e) by step scanning at 0.02 degrees 2e with a counting time of

0.5 seconds. Patterns obtained were compared with JCPDS data for

standard patterns of all probable structures.

A second specimen from the middle ribbon section was examined

optically for ribbon contact pattern and a 50x photo-micrograph

recorded for each test run. In addition, ribbon section pairs from

the first order tests involving extreme values of each independent

23

variable (Vs,

Pe, Gs) were examined by SEM, and 120x photo-micrographs

of the cooling wheel contact pattern recorded.

Finally the middle ribbon section from each experimental run was

examined visually for dimensional uniformity and edge quality and

evaluated qualitatively in comparison with commercial Metglas 2605 on

a scale of 0 (poor) to 5 (excellent). Figure 11 shows selected ribbon

specimens representative of the qualitative ratings of 0 to 5. Ribbon

specimens produced during experimental process runs were evaluated

against these standard specimens to obtain a ribbon quality number.

24

Figure 11. Ribbon Specimens Used to Determine Qualitative Ratings for DimensionalUniformity

25

4.0 RESULTS

Results of the 28 experimental runs conducted for the statistical

characterization program are presented in Tables II and III.

Table II lists results for the 15 test runs conducted for determination

of the single process variable effects. For each run, experimental

values of the three process variables are listed along with the

average measured ribbon thickness of the middle ribbon section, and

qualitative evaluations of ribbon structure and dimensional quality.

Evaluation techniques used to determine the qualitative structural

classification of the ribbon specimens, listed as the peak intensity

ratio (P.I.R.), is discussed in Section 4.3.

The results for the 13 interaction experimental runs are listed

in a similar manner in Table III.

Calculations of the average percent difference in S between the

extreme ribbon sections and middle, and corresponding standard

deviation for each are listed in Table TV. Although the average

percent difference in thickness between middle and final ribbon

sections is within measurement errors, the initial 0.5 m of ribbon

was often found to be significantly thicker (16.6%) than the middle

ribbon section.

4.1 REGRESSION ANALYSIS OF SINGLE VARIABLE EFFECTS ON RIBBON

THICKNESS

The results for the 15 single variable runs were analyzed using

the Hewlett-Packard single variable regression and VZComp programs on

an HP-87 microcomputer. For each independent process variable (Vs,

Pe, Gs ), best fit curves were plotted for ribbon thickness on an PP

x-y plotter, using the VZCury curve fitting program. Of the three

process variables, the ejection pressure was found to have the most

linear effect on ribbon thickness. Although a power curve fit

resulted in a slightly higher coefficient of determination

Table II

Results of Single Variable Experimental Runs

PC = Partially Crystalline, PA = Primarily Amorphous

Run

No.

Process Variable

Vs

Pe

Gs

(m/s) (KPa) (mm)

Nozzle Slot

Width (Wn)

(mm)

Avg. Ribbon

Thickness (S)

(pm)

Dimensional

Quality

Structure/

P.I.R. No.

1B 17.5 10.3 0.229 0.80 29.7 3 PC / 0.3404B 17.5 10.3 0.229 0.82 27.9 3 PC / 0.3487 17.5 10.3 0.229 0.85 31.2 3 PA / 0.0407B 17.5 10.3 0.229 0.65 29.2 3 PA / 0.03225 17.5 10.3 0.229 0.85 28.5 3 No x-ray

2A 17.5 10.3 0.305 0.85 37.7 3 PC / 0.0803A 17.5 10.3 0.152 0.85 26.5 3 PA / 0.04020 17.5 10.3 0.381 0.75 44.6 4 PA / 0.048

5 17.5 6.9 0.229 0.82 ?3.6 1 PC / 1.006 17.5 13.8 0.229 0.82 36.6 4 PC / 0.05221 17.5 17.2 0.229 0.65 39.0 4 Amorphous

8 22.0 10.3 0.229 0.65 28.5 2 PC / 0.0569 13.0 10.3 0.229 0.82 36.1 4 PC / 0.50018 23.6 10.3 0.229 0.75 24.3 2 Amorphous19 10.0 10.3 0.229 0.75 42.6 3 PC / 0.660

Table III

Results of Interaction Effects Experimental Runs

PC = Partially Crystalline, PA = Primarily Amorphous

Run

No.

Process Variable

Vs

Pe

Gs

(m/s) (KPa) (mm)

Nozzle Slot

Width (Wn)

(mm)

Avg. Ribbon

Thickness (S)

(pm)

Dimensional

Quality

Structure/

P.I.R. No.

10 22.0 7.0 0.305 0.65 28.7 1 PC / 0.18011 13.0 6.9 0.305 0.65 41.4 1 PC / 0.84012 22.0 13.8 0.305 0.65 37.2 4 PC / 0.060

13 13.0 13.8 0.305 0.72 56.8 3 PC / 0.35214 22.0 6.9 0.152 0.72 23.0 0 PC / 0.06415 13.0 6.9 0.152 0.70 31.8 1 PC / 0.500

16 22.0 13.8 0.152 0.72 25.8 3 PA / 0.02017 13.0 13.9 0.152 0.72 41.5 4 PA / 0.02822 13.0 13.8 0.152 0.65 37.2 4 No x-ray

23 22.0 13.8 0.305 0.65 38.6 3 No x-ray24 22.0 13.8 0.229 0.75 25.2 3 Amorphous26 22.0 13.8 0.152 0.85 24.3 3 No x-ray

27 13.0 6.9 0.305 0.85 42.5 1 No x-ray

28

Table IV

Average Percent Variation and Standard Deviation of Thickness

Over the Length of Ribbon Specimens

Average %AS

Standard Deviation

Ribbon Section

Start-Middle Middle-Finish

16.6% -1.02%

9.2 2.85

(R2 = 0.942), the difference between it and the linear fit coefficient

of determination (R2 = 0.935) was not considered significant.

The plotted curve of pressure vs thickness is shown in Figure 12

for the linear regression equation. As Figure 12 indicates, ribbon

thickness is found to increase as the ejection pressure, Pe, increases.

Thus for constant values of Vs

and Gs

, increasing the ejection pressure

results in an increased melt flow rate to the cooling wheel.

The effect of cooling wheel surface velocity, V,, on ribbon

thickness was found to be significantly non-linear over the range of Vs

values used in the experimental runs. Single variable regression

analysis of the nine data sets in which Vs was varied indicated that

ribbon thickness decreases in a non-linear manner with increasing

cooling wheel surface velocity. The power regression fit for Vs vs S

resulted in a correlation value of R2 = 0.912 and the plot of the power

fit regression is presented in Figure 13.

Although the functional relationship between ribbon thickness and

crucible nozzle clearance gap (Gs) can be represented linearly with good

accuracy, an exponential relationship results in a higher correlation

coefficient of R2 = 0.923. The plot of the exponential curve for ribbon

thickness as a function of crucible nozzle gap appears in Figure 14.

50E

U)U)

a)

40U_

CO0

30

Ejection Pressure P. (kPa)

16 18 20

Figure 12. Linear Curve Fit for Ribbon Thickness (S) as a Function of Crucible Ejection Pressure (Pe)

10 12 14 16 18

Roll Speed Vs (m/s)

20 22 24

Figure 13. Linear Regression Power Fit Curve for Ribbon Thickness (S) as a Function of CoolingRoll Speed (V$)

60

50

40

30

200.15 0.20

S(Gs) = 16.91 e2.504 (Gs)

0.25 0.30

Nozzle Gap Gs (mm)

0.35 0.40

Figure 14. Exponential Curve Fit for Ribbon Thickness (S) as a Function of Crucible NozzleClearance Gap (Gs)

32

For all equations presented for thickness, the following units were

used to obtain S in cooling wheel velocity (Vs) in m/s, crucible

ejection pressure (Pe) in KPa, and crucible nozzle clearance gap (Gs) in

mm. The three single variable equations for ribbon thickness as a

function of the independent process variable are listed below:

S(Pe) = 13.26 + 1.56 (Pe) (2)

S(Vs) = 166.9 (Vs)-0.6014

(3)

S(Gs) = 16.91e2'504(Gs)(4)

4.2 RESULTS OF MULTIPLE LINEAR REGRESSION ANALYSIS

Multiple linear regression (MLR) analysis was performed on the

results of the 28 experimental runs using a Hewlett-Packard HP-87

microcomputer and HP-85 Statistic Pack MLR program. For each test run,

operating values for 1c, andand Gs were entered as independent

variables and the measured thickness of the middle ribbon section as the

dependent variable. The MLR program was first run for all ten first

order, second, and interaction terms. The MLR coefficients and

corresponding F and T values were printed out in addition to the R2

value. Analysis of the results indicated that the (Pe)2, (Vs) x (Gs),

and (Vs

) x (Pe

) x (Gs

) terms were relatively insignificant. After

eliminating these terms from the analysis, a second MLR analysis was

performed resulting in an equation for ribbon thickness involving a

constant and seven variable terms. The multiple coefficient of

determination (R2) for the seven-term MLR equation is 0.926 as compared

to R2 = 0.937 for the 10-term MLR equation. The MLR analysis results

for the seven variable ribbon thickness model are listed in Table AI of

the Appendix A and include F values for the variables and corresponding

T values for the coefficients. The resulting equation for ribbon

thickness, S, in um as a function of cooling wheel surface velocity in

m/s, crucible ejection pressure in kPa, and nozzle clearance gap G, in

mm is given in Equation (5).

S = 72.80 - 4.10 (Vs) + 1.45 (Pe) - 144.8 (Gs)

+ 0.10(V5)2 + 346.0(Gs)2

- 0.08(Vs)*(Pe) + 5.32(Pe)*(Gs)

33

(5)

4.3 RESULTS OF X-RAY DIFFRACTION ANALYSIS OF RIBBON STRUCTURES

X-ray diffraction scans of the ribbon cooling wheel surfaces were

made for 23 of the experimentally cast ribbons. Diffraction scans were

also performed on the ribbon free (top) surface for six of the ribbon

specimens. Complete diffraction scans from 5 degrees to 90 degrees, 2e,

are included in Appendix B.

The diffraction patterns for all ribbon specimens studied show a

broad amorphous halo at the approximate location of the a-Fe (110) peak.

Ribbon specimens from experimental runs 18, 21, and 24 were found to

have no detectable crystalline peaks over the range of the diffraction

scans. The remaining ribbon specimens were found to have detectable

crystalline peaks at the approximate location of the (200) peak of a-Fe.

In addition, several ribbon specimens studied were found to have

diffraction peaks at the location of the amorphous halo.

The diffraction peaks observed at the approximate location of the

(200) a-Fe peak varied in intensity from 0.5 x 102 counts for run 16, to

a maximum of 2.5 x 102 counts for Run 5. The maximum peak intensity,

2.5 x 103 counts was used to calculate the normalized peak intensity

ratios (P.I.R.) listed for the Experimental runs in Tables II and III.

This ratio, which can be interpreted only in a aualitative manner, was

used to divide the ribbon specimens arbitrarily into three structural

groups: partially crystalline (0.05<PIR<1.00); primarily glass

(0.01<PIR<0.05); and amorphous (no detectable crystalline peaks).

Typical examples of x-ray diffraction patterns obtained for each of the

structural categories are shown in Figure 15.

(a) Amorphous

10.0

8.0

6.0

4.0

2.0

0.0

r-

I

10.0 26.0(b) Primarily Amorphous (P.A.)

15.0

X = (200) Peak of Alpha Iron12.0 -

0

42.0

9.0

o 6.0

3.0

0.0

58.0 74.0 90.0

5.0 22.0

(c) Primarily Crystalline (P.C.)15.0

12.0

9.0

6.0

3.0

0.0

39.0

X = (200) Peak of Alpha Iron

5.0

56.0

X

73.0 90.0

22.0 39.0 56.0Degrees 20

Figure 15. Examples of X-ray Diffraction Patterns Obtained forEach of the Structural Groups.

73.0 90.0

34

35

Two ribbon specimens were selected from each of the three

structural categories and submitted for x-ray diffraction analysis of

the free surface. The diffraction results for both surfaces of the six

ribbon specimens thickness are presented in Table V.

Table V

X-Ray Diffraction Results for Cooling Wheel and Free Surfaces

Of Selected Ribbon Specimens

StructuralClassification

RSCRun No.

Cooling Whl.Surface (P.I.R.)

FreeSurface (P.I.R.)

Amorphous18

24

No Peaks

No Peaks

No Peaks

No Peaks

Primarily 3A 0.040 No Peaks

Amorphous 16 0.020 No Peaks

Partially 5 1.00 0.124

Crystalline13 0.352 No Peaks

4.4 RESULTS OF QUALITATIVE EVALUATION OF RIBBON DIMENSIONAL QUALITY

The mid-section ribbon specimens from each exprimental run was

examined visually for uniformity of width and surface appearance, and

discontinuities along the ribbon edges. The results of the qualitative

ratings for the 28 experimental runs are given in Tables II and III.

The ratings assigned to the ribbon specimens ranged from 0 (very poor)

for run 14 to 4 (very good) with a number of runs. The majority of

ribbon specimens were rated good (3) for dimension quality.

36

Although several of the ribbon specimens displayed a surface appearance

comparable to that of the commercial materials, in general they did not

have comparably smooth edges.

4.5 RESULTS OF RIBBON COOLING WHEEL SURFACE CONTACT PATTERN STUDIES

The cooling wheel contact pattern of ribbon specimens from each of

the 28 experimental runs was studied optically, and photo-micrographs of

typical contact patterns recorded for each specimen. In addition,

experimental runs involving extreme values of the variables Vs, Pe, and

Gswere studied using SEM techniques to determine the effects of large

process variable changes on the ribbon contact surface pattern.

37

5.0 DISCUSSION OF RESULTS

5.1 EFFECTS OF PROCESS VARIABLES ON RIBBON THICKNESS

The thickness equation (Equation (5)) developed from MLR analysis

provides a relatively simple empirical model for determination of

thickness as a function of the three controllable process variables, Vs,

Pe, and Gs. In contrast to the empirical relationships developed from

volume flow rate or high speed photographic measurements, which relate

thickness indirectly to the process variables, the MLR equation provides

a method for calculation of thickness directly from the controllable

process variables.

The relationship presented for thickness, S, involves a linear

combination of first and second order terms for Vs, Pe, and G. From an

examination of the equation, it is found that Vs and Gs are present as

both first and second order terms, while Pe is found only as a first

order variable.

In addition, significant interaction terms exist between both roll

speed and crucible ejection pressure, and nozzle clearance gap and

crucible ejection pressure. This indicates that the response of

thickness to changes in an individual process varable will depend not

only on the operating range of that variable but also on the operating

values of the two process variables being held constant.

As indicated previously, Pewas found to behave in a linear manner

throughout the experimental range. In contrast, Huang(o) found that S

was proportional to the /Pe for ejection pressures between 5.5 kPa and

20 kPa. His relationship, which was derived from the Bernoulli

equation, does not take into account melt-flow constraints due to the

presence of the cooling wheel surface. While Huang found that the

experimental data follow the relationship S./Pe for nozzle slot

width-to-gap ratios around unity (Wr = Gs), the equation did not hold

true for Wn/G

s>1. Comparison of casting conditions for the single

variable experimental runs with those employed by Huang show that

38

although the range of Pe and the values for Vs are similar, the Wn/Gs

ratio for the present work is between 2.8 and 3.7 as compared to a

Wn/G

sratio of around 1.0 for Huang's work. Huang attributed the

dependence of ribbon thickness on the nozzle slot width to gap ratio

to increased constraint of the melt-flow as Wnbecomes increasingly

greater than Gs. It is of interest to note that Huang found that as

Wn/G

sincreases above unity, ribbon thickness is less than predicted

by the S.= 1/13e relationship. This is in agreement with the present

results.

Analysis of the experimental error for thickness based on

Equation 5 is included in Appendix A. The results indicate an error

of approximately 6.8% for ribbon thickness. Although this value is in

general agreement with observed thickness variations between repeated

process runs, it should be noted that the estimated measurement error

for Gsused in the calculations does not include the eccentricity of

the cooling wheel. If the value for cooling wheel runout is included

in the estimated error for Gs

, the calculated error for thickness, S,

becomes unrealistically large. This observation, which is in

agreement with the discussion of process operation presented by

Narasimhan(1) indicates that ribbon thickness is somewhat insensitive

to variations in nozzle clearance gap.

While the MLR equation developed from the experimental data

provides a convenient method for predicting the effects of the

controllable process variables on ribbon thickness, certain restrictions

must be considered for its use.

First, the range of variable values for Vs, Pe, and Gs was

limited by machine capabilities. In particular, the cooling wheel

velocity was limited to a maximum of 23.6 m/s because of the drive

ratio selected for the electric motor. While the ejection pressure

and nozzle gap were not limited by machine capabilities, large

increases in either variable without a corresponding increase in

cooling wheel speed would result in non-optimum ribbon thickness.

39

A second restriction exists for application of the present

results to other PFC process equipment configurations and operating

conditions. The MLR equation presented here represents the best

regression fit for the 28 experimental runs conducted for the

characterization program. As illustrated in the previous comparison

of the effects of ejection pressure, differences in operating

conditions can lead to significantly different relationships between

thickness and the independent variables.

In a similar manner, major variations in machine configuration

can be expected to change the functional relationship between

thickness and the process variables. Therefore, use of the MLR

equation for calculation of thickness under different operating

conditions or machine configurations may result in significant errors.

5.2 EFFECT OF PROCESS VARIABLES ON RIBBON COOLING SURFACE PATTERN

The type of contact obtained between the cast ribbon and the

cooling wheel surface appears to be a major factor in determining

ribbon structure. Because gas pockets represent regions of poor local

heat transfer, selection of process variable operating conditions

which minimize both the number and size of the gas pockets is an

important consideration in optimization of the PFC process for

production of amorphous ribbon.

Optical and SEM studies conducted on the cast ribbon specimens

revealed considerable variation in cooling surface contact patterns.

As was expected, those ribbons exhibiting uniform distributions of

small gas pockets in general had structures which were amorphous, or

primarily amorphous. Ribbons having contact patterns characterized by

numerous, large, elongated gas pockets showed higher decrees of

crystallinity.

Evaluation of the effects of the process variables on the ribbon

contact pattern was made on the basis of paired studies involving

extreme operating values for Vs, Pe, and Gs

. For each comparison, the

40

other two controllable process variables were held constant at the

mid-values. The contact pattern of commercial Metglas 2605 was used

for comparison with the ribbons cast in the present study.

The optical photomicrograph of the MG2605 ribbon cooling wheel

contact surface appears in Figure 16 while the SEM photograph of the

same contact pattern is presented in Figure 17.

5.2.1 Effect of Cooling Wheel Surface Velocity (Vs)

Experimental runs were performed at low and high values of Vs,

while medium operating values for Pe and Gs were used for both runs.

Optical micrographs of the ribbon contact pattern obtained at Vs =

23.6 m/s (Specimen 18) and Vs = 9.97 m/s (Specimen 19) are shown in

Figures 18 and 19 respectively. The contact pattern for high Vs

(Figure 18) exhibits a relatively uniform distribution of small gas

pockets and several larger, randomly distributed gas pockets. Areas of

contact between the ribbon and cooling wheel show good replication of

the sanded cooling wheel surface. This sample, which was found to be

amorphous by x-ray diffraction analysis, appears to be reprsentative of

the contact pattern necessary for high overall cooling rates. Although

the cooling wheel surface condition used in commercial production of the

Metglas product is not known, comparison of specimen 18 with the optical

(Figure 16) and SEM (Figure 17) micrographs for MG2605 shows a similar

gas pocket morphology.

In contrast, Specimen 19, produced at Vs = 9.97 m/s, displays a

very poor ribbon contact pattern characterized by an almost continuous

network of gas pockets and channels. The optical micrograph of the

contact surface (Figure 19) shows only a few, localized bands in which

any contact between the ribbon and cooling surface appear to have

occurred. The SEM micrographs for ribbon contact patterns produced at

high Vs and low Vs are shown in Figures 20 and 21 respectively.

41

At:Vg205

Figure 16. Cooling wheel surface contact pattern for commercial metallicglass (MG 2605), 50X.

Figure 17. SEM photograph of cooling wheel surface contact pattern forcommercial metallic glass (MG 2605), 120X.

a)

C)Li=

a)

IL

43

100U 001 00018 BOM

Figure 20. SEM photograph of ribbon cooling wheel contact surfaceproduced at high Vs (23.6 m/s, 120X.

20KV X120 100U 002 00019 BOMFigure 21. SEM photograph of ribbon cooling wheel contact surface

produced at low Vs (9.97 m/s, 120X.

44

This comparison indicates that higher cooling wheel velocity, in

addition to reducing ribbon thickness, is a major factor in reducing the

average gas pocket size and improving contact between the ribbon and

cooling wheel surface.

Huang and Fiedler(4)

reported similar results in a study made on

the effects of cooling surface velocity on amorphous ribbon formation.

In their study, the metal-wheel interfacial conductance, h, was found to

increase from 12 cal/cm2sK at V

s= 11.7 m/s to 50 cal/cm

2sK at

Vs = 26.6 m/s, as estimated from a continuous casting model. This

increase in h was believed to contribute to improvements in the ribbon

cooling surface contact pattern at higher cooling substance velocities.

Measurements of percent area of ribbon-cooling wheel contact varied in a

like manner; from 50% at Vs = 11.7 m/s to 85% of the total area at

Vs

= 26.6 m/s. Comparison of the SEM photographs in Figures 20 and 21

with SEM photographs presented by Huang and Fiedler shows a similar

relationship between contact area and morphology, and cooling wheel

surface velocity.

5.2.2 Effect of Ejection Pressure (Pe)

The optical micrographs of the ribbon contact pattern for experi-

mental runs involving extreme values of Pe appear in Figure 22 for

Pe= 17.2 KPa (Specimen 21) and Figure 23 for P

e= 6.9 KPa (Specimen 5).

Comparison of the two micrographs demonstrates the effectiveness of

increased ejection pressure at reducing the size and area of gas pockets

Whereas the gas pockets present in Specimen 5 (Pe = 6.9 KPa) are

typically 200-400 pm in length and up to 50 pm in width, those present

in specimen 21 (Pe = 17.2 KPa) generally do not exceed 50 pm length and

10 pm width. In addition, areas of contact between the ribbon and

cooling wheel are characterized by networks of gas pocket channels for

the low pressure run while contact areas produced at the high ejection

pressure more nearly duplicate the cooling wheel surface features.

45

Figure 22. Ribbon cooling wheel contact surface produced at high Pe(17.2 KPa), 50X.

Figure 23. Ribbon cooling wheel contact surface produced at low Pe(6.9 KPa), 50X.

46

The SEM micrographs of the contact pattern for both runs are shown

in Figures 24 and 25. The contact pattern produced at the high ejection

pressure appears in Figure 24 while the contact pattern for the low

pressure run is shown in Figure 25. These photographs show more clearly

the difference in areas of contact between the ribbon and cooling wheel.

The areas of contact observed for high Pe are quite similar to those

found in the MG 2605 specimen while those produced at low Pe appear

smoother, indicating less intimate contact between ribbon and cooling

wheel surface.

These observations appear to be supported by the x-ray diffraction

results. Specimen 5, cast at a low Pe value, was found to have the

strongest crystalline peak of all ribbon specimens analyzed. Results

for Specimen 21 indicate the ribbon is amorphous.

These qualitative results are in general agreement with the study

of melt delivery conditions reported by Huang.(8)

Although the

processing conditions reported by Huang differ somewhat from those in

the present work, the SEM photograph of the ribbon contact pattern

presented by Huang for low ejection pressure (9.0 KPa) shows large,

elongated gas pockets similar to those present in Figure 25. While a

comparison of SEM photographs of the ribbon contact patterns obtained at

high ejection pressures reveals a similar trend toward smaller, more

uniformly distributed gas pockets, Huang's work involved considerably

higher ejection pressure values (25 KPa) than those used in the present

study (17.2 KPa). In addition, all casting reported by Huang was

conducted with a nozzle clearance gap of 0.35 mm and cooling wheel

surface velocity of 15.0 m/s as compared with the present work in which

Gs

and Vswere 0.229 mm and 17.5 m/s respectively.

5.2.3 Effect of Nozzle Clearance Gap (Gs)

In contrast to both Vs

and Pe

, the nozzle clearance gap value

appears to have a relatively minor effect on the ribbon contact pattern.

Optical micrographs of the contact surfaces appear in Figures 26 and 27

for Gs= 0.381 mm (Specimen 20) and G

s= 0.152 mm (Specimen 3A),

47

Figure 24. SEM photograph of ribbon cooling wheel contact surfaceproduced at high Pe (17.2 KPa, 120X.

100U 011 00005 BUM

Figure 25. SEM photograph of ribbon cooling wheel contact surfaceproduced at low Pe (6.9 KPa, 120X.

tf)

.00)

CD

UN

70aa)

co

U)

LO

C8a)tv

_c3Xer)c

EO ECo c;_o1:2 c")

EE 6

C.1

a)

DO)

LE

/CI

UN

7-100a0CD

LO4-

15U)

CO

CU

,T)a)

3cr) 0c

O EU EC fVO Lo_0-0O

cva)1

0)

LL

49

respectively. Although the contact pattern for Specimen 3A (Gs =

0.152 mm) exhibits a number of relatively large gas pockets when

compared to Specimen 20 (Gs = 0.381 mm), this would appear to be due

primarily to the difference in ribbon thickness for the two samples.

Thin ribbons less than 30 um produced at low and medium ejection

pressure values typically have a periodic or "fish scale" gas pocket

pattern. Under such conditions, the ribbons tend to have periodic

regions of larger gas pockets separated by regions having smaller,

less numerous gas pockets. Comparison of the SEM micrographs in

Figures 28 and 29 shows very similar ribbon contact patterns for

Specimen 20 (Figure 28) and Specimen 3A (Figure 29). The absence of

large elongated gas pockets in the SEM photograph of Specimen 3A is

most likely due to the more limited field of view available at the

higher magnification.

The relatively minor differences in ribbon contact pattern

observed for Specimens 3A and 20 appear to have little or no effect on

ribbon structure. X-ray diffraction analysis of ribbon samples from

the two runs resulted in approximately equal values of 0.040 for

Run 3A and 0.048 for Run 20. Both were classified as primarily

amorphous.

50

Figure 28. SEM photograph of ribbon cooling wheel contact surfaceproduced at high Gs (0.382 mm, 120X.

Figure 29. SEM photograph of ribbon cooling wheel contact surfaceproduced at low Gs (0.152 mm, 120X.

51

6.0 CONCLUSIONS

1. Rapidly solidified ribbons with good dimensional uniformity and

appearance were produced in small batch quantities utilizing a

relatively simple, inexpensive planar flow casting machine.

2. Analysis of single variable process runs, designed to study the

effects of the three controllable process variables on ribbon

thickness, indicates that although the ejection pressure, Pe,

affects thickness in a linear manner, both the cooling wheel

surface velocity, Vs, and the crucible nozzle clearance gap, G5,

have a significantly non-linear effect on thickness.

3. Unlike previous studies of the PFC process which have considered

the effects of a single variable on process operation, the use of

multiple linear regression analysis techniques enables the

effects of several process variables to be included in a single

empirical equation for thickness.

4. MLR analysis of 28 experimental process runs resulted in an

equation for thickness which includes both the non-linearities as

well as the product terms for the three controllable process

variables, Vs, Pe, and Gs

. This analytical approach allows the

prediction of the effect of simultaneous changes in process

variable values on the ribbon thickness.

5. Increases in either Vsor P

e, or both, up to the values used in

this study produced significant improvements in the ribbon

contact pattern morphology. Ribbon specimens produced at high

values of Vs

and Pedisplayed an increased degree of

amorphousness as well as improved surface uniformity.

52

6. Although effective at characterizing the operating response of

the PFC machine, the statistically designed experimental approach

does not always allow direct comparison of results with relation-

ships developed from more conventionally designed experiments.

53

BIBLIOGRAPHY

(1) Narasimhan, M. C., US Patent No. 4 142 571, (1979).

(2) Kavesh, S., "Metallic Glasses," American Soc. Metals, Metals Park,Ohio, (1978).

(3) Vincent, J. H.; Herbertson, J. G.; Davies, H. A., Proc. 4th Int.Conf. on Rapidly Quenched Metals (Sendai, 1981), 77-80.

(4) Huang, S. C.; Fiedler, H. C., Mater. Sci. and Eng., 51 (1981),39-46.

(5) Liebermann, H. H., Mater. Sci. and Eng., 43 (1980), 203-210.

(6) Hillmann, H.; Hilzinger, H. R., Proc. Rapidly Quenched Metals III,Vol. 1, Metals Society, London (1978), p. 22-29.

(7) Vincent, J. H.; Davies, H. A.; Herbertson, J. G., "Cont. CastingSmall Cross Sections," Fall meeting TMS-AIME (1980).

(8) Huang, S. C., Proc. 4th Int. Conf. on Rapidly Quenched Metals,(Sendai, 1981), p 65-68.

Huang, S. C.; Fiedler, H. C., Metall. Trans. A, Vol. 12A,June 1981, p 1107-1112.

(10) Johnson, L. A., (General Electric, USA), personal communication(1982) (9/15/82).

(11) Pavuna, D., J., Non-Cryst. Sol., 37 (1980), 133-137.

(12) Pavuna, D., J., Mater. Sci., 16 (1981), p. 2419-2433.

(13) Liebermann, H. H., Proc. Rapidly Quenched Metals III, Vol. 1,Metals Society, London, (1978), p. 34-40.

(14) Seybolt, A. U.; Burke, J. E., "Procedures in ExperimentalMetallurgy," 1969.

(15) Decristofaro, N. F.; Freilich, A., J. Mater. Sci. 17, (1982),p. 2365-2370.

(9)

APPENDICES

54

APPENDIX A

A.1 - RESULTS OF MULTIPLE LINEAR REGRESSION ANALYSIS

OF PFC PROCESS RUNS

The multiple linear regression analysis coefficients determined for

28 experimental PFC process runs are presented in Table AI. Also

included are corresponding F and T values determined for the seven-term

MLR Equation (5).

Table AI

Results of Multiple Linear Regression Analysis

for 28 PFC Process Runs

Variable Term Coefficient B (I) F Value T Value

Const. 72.80

Vs

-4.10 107.40 -3.08

Pe

1.45 31.05 1.32

Gs

-144.80 91.63 -2.20

(Vs)2 0.10 6.31 2.75

(Gs)2 346.02 9.02 3.13

(Vs)*(Pe) -0.08 3.16 -1.72

(Pe)*(Gs)5.32 3.38 1.84

Multiple Correlation

Coefficient R2= 0.926

55

APPENDIX A

A.2 - ERROR ANALYSIS OF SEVEN VARIABLE MLR EQUATION

FOR RIBBON THICKNESS (S)

The 7 variable MLR equation for ribbon thickness, S, as a function

of the three process variables Vs, Pe, and Gs

is:

S = 72.80 - 4.10(Vs) + 1.45(Pe) - 144.8(Gs)

+ 0.10(Vs)2 + 346.0(Gs)2 - 0.08(Vs)*(Pe)

+ 5.32 (Pe)*(Gs)

(5)

Substituting the quantities Vs + AVs, Pe + APe, and Gs

+ AGs

into

Equation (5) and subtracting it from the original equation for S results

in the following equation for AS after higher order terms are eliminated.

AS = 4.10 (AVs) - 1.45(APe) + 144.8(AGs)

- 0.10(2Vs*AVs) - 346.0(2Gs*AGs)

+ 0.08(Vs*APe + AVs*Pe) - 5.32(Pe*AGs + APe*Gs)

(6)

Estimating the maximum error for each of the process variables as

one-half the smallest readable division on each of the measuring

instruments, we have:

AVs

= ± 0.2 m/s

APe

= ± 0.2 KPa

AGs

= ± 0.03 mm

Calculating the percent error in S for the mid-value process runs

(Vs

= 17.5, Pe

= 10.3, and Gs= 0.229) gives the following error

estimate:

AS _ 2.023x 100 g ow

s--A ILIV = UsUM

56

APPENDIX B

B.1 - X-RAY DIFFRACTION PATTERNS OF EXPERIMENTALLY CAST

RIBBON SPECIMENS: COOLING WHEEL SURFACE

RSC- IB SOT TOM SURFACE

ALPHA IRON

10.0 26.0 42'.O 58.0 74.0 90.0

15.0 RSC-2A; MG RIBBON X. (209) PEAK OFALPHA IRON

100

BO

60

40

29

5.0 22.0

39.0 56.0

39.0 56.0

73.0 90.0

ALPHA IRON696

73.0 90.0

15.0

12.0

9.0

6.0

3.0

RSC-3A; MG RIBBON X.= (200) PEAK OFALPHA IRON

5.0

100

80

60

40

20 1.

5.0

22.0

22.0

39.0 56.0 73.0

39.0 56.0

90.0

ALPHA IRONS- 696

73.0 90.0

fl

4.0

RSC-49; BOTTOM SURFACE

lumowles4NOwoo4,14,0iA

in.n 26.0 42.0 58.0 74.0 qn.0

100

80

(ILPHA IRON

-20

10.0 26'.0 42.0 581.0 74'. 0 clA.0

.(.1 RSC-5; MG RIBBON(NOTE: SCALE 2X THAT

A

I

1.2

a

0.6

OF OTHER SCANS)

5.0

100

80

60

40

20

5.0

22.0 39.0

22.0 39.0

in3

56.0

x

X. (200) PEAK OFALPHA IRON

73.0 90.0

ALPHA IRON6. 696

56.0 73.0 90.0

15.0 RSC-6 MG RIBBON X.= (200) PEAK OF ALPHAIRON

100

80

60

40

20

h. ALPHA IRON5- 696

5.0 22.0 39.0 56.0 73.0 90.0 01N

15.0 RSC-7 MG RIBBON X.= (200) PEAK OFALPHA IRON

100

80

60

40

20

39.0 56.0 73.0 90.0

ALPHA IRON696

5.0 22.0 39.0 56.0 73.0 90.0

10. RSC -7B; COOLING WHEEL X.=(.200) PEAK OFALPHA IRON

6.0

4.0

2,A

-f10.0 26.0 42.0 58.0 74.0 90.0

10n

80

60

40 ALPHA IRON6- 696

10 .A 26.0 rI

58.0 74.0 90.0

10 (3

4 .

-RSC-13; COOLING WHEEL X.= (200) PEAK OFALPHA IRON

10.0

100

80

60 -

40

20

26.0

10..0 26.0

42.0 58.0

42.0 58.0

74.0 90.0

ALPHA IRON6- 696

74.0 90.0

15. RSC-9; MG RIBBON X.= (200) PEAK OFALPHA IRON

100

80

60

40

20

ALPHA IRON696

5.0 22.0 39.0 56.0 73.0 90.0

6.n

6.0

4.0

-RSC-i& COOLING WHEEL X.= (200) PEAK OFALPHA IRON

Y.= <110) PEAK OFALPHA IRON

10.0

I no

en -

60

40

20

10.

26.0

26.0

42.0 58.0

42.0 58.0

74.0 90.0

ALPHA IRON696

74.0 90.0

. 0 RSC-11J COOLING WHEEL

1

O .8

O .4

x 163

10.0

100,.

8n .

6A .

40 .

20

cb..'"Ik,"0 42

1

.0 58.0

X.= (200) PEAKY.= (110) PEAK f- OF

ALPHA IRON

74.0 90.0

ALPHA IRON6- 696

10..0 26.0 42.0 58.0 74.0 90.0

J0.0 -RSC-12 COOLING WHEEL

8.0

6.0

X.= <200) PEAK OFALPHA IRON

1E . 0

100

SO

6n

40

20

ALPHA IRON6- 696

10..0 26.0 42.0 58.0 74.0 90.0

.RSC-13; COOLING WHEEL X.= (200) FEWY. (110) PEAK

ALPHA IRON

10.0

100 -

80

60

40

20

26.0 42.0 58.0 74.0

ALPHA IRON6- 696

10.6 26.0 42.0 58.0 74.0 90.0

OF

G.0

4.0

p

RSC-14; BOTTOM SURFACE

1n0

60

40

20

26.0I

42.0I

56.0I

74.0

ALPHA IRON6- 6-7.4;

10.0 26.0 42.0 58.0 74.0 90.0

2 0 -RSC-15; COOLING WHEEL

1.6

1

0.8

0.4

X..= (200) PEAK ))

..--1101EOPEAIALPHA IRON

10.0

100

30a

60

40

20

26.0

10..0 26.0

42.0 58.0 74.0 30.0

ALPHA IRON6 696

42.0 58.0 74.0 90.0

15.0 RSC-16; MG FE-8-S I X.= (200) PEAK OFWHEEL SURFACE

12.0

9.0

6.0

3.0

ALPHA IRON

5.0

100

80

60

400

20

22.0 39.0 56.0 73.0 90.0

ALPHA IRON696

5.0 22.0 39.0 56.0 73.0 90.0

10.0 RSC-17; COOLING WHEEL

8.0

X.= (200) PEAK OFALPHA IRON

60

40

20

.0 42.0 58.0 74.0 90.0

10,0 26.0 42.0 58.0

ALPHA IRON696

74.0 90.0

10.0 -RSC-18; COOLING WHEEL

8.0

6.0

4.0

2.0

10.0

100 -

80

60 r

40

20

26.0

10.0 26.0

42.0 58.0

42.0 58.0

74.0 90.0

ALPHA IRON6- 696

74.0 90.0

1.6

A

n

13.4

RSC-19 BOTTOM SURFACE X 107

10.0

100 -

80

60-413

20

26.0 42.0 58.0 74.0

10.0 421 .0 58'.Or

74.-0

ALPHA IRONr.- 69F.

901.13

4.0

4

RSC-20; BOTTOM SURFACE

10.0

100

80

60

40

'70

26.A 42 0 58.0 74.0

ALPHA IRONGc.4=.

10.0 26.0 42.0 58.0 74.0 90.0

10.e. -RSC-21; BOTTOM SURFACE

4 . 0

E30

40

2 G1

1 0 . 0 26.0 4a.0 58.0

ALPHA IRON69r.

74.0 90

-RSC-24; BOTTOM SURFACE

z, t

4 CI

10.0 26.0 42.0 58.0 74.0 90.0